Issue link: https://iconnect007.uberflip.com/i/474043
42 The PCB Magazine • March 2015 ture and the higher the peak testing tempera- ture is above the laminate T g the more uncer- tainty there is about the results as the m con- stant may be significantly underestimated. For thermal testing above T g the model has a strain component, and below T g thermal cycling the model accounts for strain ratcheting, with the m constant being empirically derived. For thermal cycling temperatures below ~140°C, empirical work would be needed to determine the m constant. The model was built around laminates in the 170–200°C T g and may not be appropriate for other laminate T g ranges. How- ever, the goal is not to determine if a particular design is reliable, it's simply to equate lab test- ing to field use conditions. Humidity was not considered in this model, and it's well known that moisture within the laminate will volatil- ize during thermal excursions. It's assumed that the thermal cycling test coupons are handled and stored properly before testing; industry- recognized standards make no mention of pre- baking test coupons [22] , but several OEMs have written their own testing protocols which in- clude pre-baking of coupons prior to thermal cycling testing. As George Box said "…all mod- els are approximations. Essentially all models are wrong, but some are useful. However, the approximate nature of the model must always be borne in mind." [23] . Once thermal cycling data is completed, failure analysis is in order. This should include distribution fitting, data exploration, reliability analysis and interpretation, and statistical pro- cess control. Conclusions Reliability testing has evolved over the last 100 years with significant improvements in ac- celerated testing methods and sophistication in data analysis; having an understanding of reli- ability terms and methods is important. Work is still needed in reaching industry consensus as to what each of the reliability test protocols prove and when each test protocol should be used. Correlating laboratory accelerated testing to field use conditions is challenging and can be misleading when test conditions are too ex- treme, as the acceleration interval becomes ex- cessively wide, which can significantly under or over estimate reliability. When thermal testing temperature approaches the failure threshold, significant variability is introduced, and larger sample sizes are recommended. Future work will focus on refining the acceleration factor and correlating accelerating testing to field use conditions. The author wishes to acknowledge and give thanks to Viola Richard, OM Group, for her work in running all the thermal cycle testing and cross section work, and Dr. Thomas P. Ryan for his review of the paper and his critiquing of the plausible Coffin-Manson fatigue model to correlate accelerated thermal cycling test condi- tions to printed circuit board assembly and field use conditions. PCB References 1. Wasserman, Gary, S. (2003). Reliability Verification, Testing, and Analysis in Engineer- ing Design. Marcel, Decker, Inc., New York, NY. 2. Zio, E. (2012). The Monte Carlo Simu- lation Method for System Reliability and Risk Analysis. Springer, New York, NY. 3. Weibull, W. (1951). "A Statistical Distri- bution Function of Wide Applicability." ASME Applied Journal of Mechanics, pp. 293–297. 4. Gray, K. No Evidence of Correlation: Field failures and Traditional Reliability Engineer- ing. No MTBF, February 10 th , 2012. Retrieved 15 Feb., 2014. 5. 2013 IPC International Technology Road- map for Electronic Interconnections. 6. Ryan, T. (2007). Modern Engineering Statistics. John Wiley & Sons, Hoboken, New Jersey. 7. Freda, M. & Barker, D. (2006). "Predict- ing Plated Through Hole Life at Assembly and in the Field from Thermal Stress Data," Proceed- ings of APEX Conference, Anaheim CA. 8. Abernethy, Robert, B. (2010). The New Weibull Handbook, 5 th Ed. Robert B. Abernethy, North Palm Beach, Florida. 9. Minitab® 16 Software, Reliability Analysis. 10. Escobar, L., & Meeker, W. (2006). "A Re- view of Accelerated Test Models." Statistical Sci- ence, Vol. 21, No. 4, 552–557. 11. O'Connor, P. & Kleyner, A. (2012). Prac- tical Reliability Engineering, 5 th Ed. John Wiley & Sons, United Kingdom. RELIABILITy TESTING AND STATISTICS continues Feature